Introduction
Within the last decades, the tremendous progress in genetic engineering paved the way for generating CAR T cells. CARs typically comprise an extracellular antigen recognition moiety fused via a flexible hinge and transmembrane region to an intracellular signaling unit, thereby combining the virtues of Abs (high antigen-binding specificity) and immune cells (potent anti-tumor effector mechanisms) within one single fusion molecule [
1]. Especially for the treatment of relapsed/refractory B-cell-derived malignancies, CAR T-cell therapy targeting the CD19 antigen has achieved remarkable clinical results [
2‐
6].
However, severe treatment-associated toxicities still restrain the widespread application of this promising technology. The most frequent side effects following CAR T-cell administration include cytokine release syndrome [
4‐
7], neurotoxicity [
6,
7], on-target/off-tumor responses [
4,
8‐
10], and anaphylaxis [
11], all of which may lead to life-threatening or even fatal implications. Given the enormously long-term persistence and proliferative capacity of genetically engineered T cells [
12], control and reversal of toxicity have emerged as important aspects of CAR T-cell therapy.
Various approaches have been pursued to mitigate side effects ranging from global, unspecific immunosuppression to selective ablation of CAR-engineered T cells. The latter strategy is currently under massive investigation and is based on the transgenic introduction of either suicide genes or elimination marker genes. Two well-studied suicide genes are herpes simplex virus thymidine kinase [
13] and inducible caspase-9 [
14‐
16] which can be triggered by the administration of small molecules to effectively induce CAR T-cell death. Alternatively, forced expression of a targetable cell-surface antigen physiologically not present on T lymphocytes has been evaluated. Proposed elimination markers are truncated EGFR [
17,
18] and CD20 [
19,
20] which are recognized by the mAbs cetuximab and rituximab, respectively. Infusion of these therapeutic molecules subsequently results in specific CAR T-cell depletion via antibody-dependent cellular cytotoxicity (ADCC) and complement-dependent cytotoxicity (CDC).
Although all of the above-mentioned methodologies allow the effective elimination of adoptively transferred T cells in case of severe toxicities, each approach exhibits distinct limitations potentially restricting a broad clinical utility. These include immunogenicity [
21], large size of the proposed depletion marker (over 130 aa), dependence on the patients’ immune system (ADCC, CDC) and occurrence of on-target adverse events due to mAb recognition of healthy tissue [
22]. Moreover, all techniques are based on the insertion of an additional gene that is co-expressed with the CAR of interest, bearing the risk of CAR T-cell escape in case the safety switch is not uniformly and/or stably expressed. Therefore, we intended to use a short peptide epitope (E-tag) directly incorporated into the CAR architecture as a targetable moiety for selective and stringent CAR T-cell elimination [
23,
24]. Another substantial drawback of currently available safeguard strategies is their reliance on pharmacological drugs whose therapeutic effect inevitably declines due to their short half-life. For that reason, we sought to use T lymphocytes as living drugs that are equipped with a CAR construct directed against a targetable portion of the therapeutic CAR. In this report, we demonstrate the utility of the E-tag as a selection marker for an efficient T-cell-mediated elimination of autologous CAR T cells.
Materials and methods
Cell lines and culture
All cell lines were kept at 37 °C in a humidified atmosphere of 5% CO
2. PC3 cells were stably transduced with the open reading frame (orf) encoding the prostate stem cell antigen (PSCA) as previously reported [
25]. The murine fibroblast cell line 3T3 and the human embryonic kidney cell line HEK293T were cultured in DMEM (ThermoFisher Scientific, Schwerte, Germany) supplemented with 10% FCS, 100 µg/ml penicillin/streptomycin and 1% non-essential aa. PC3-PSCA and Chinese hamster ovary (CHO) cells were maintained in RPMI 1640 media supplemented with 10% FCS, 100 µg/ml penicillin/streptomycin, 1% non-essential aa, 1 mM sodium pyruvate and 2 mM
N-acetyl-
l-alanyl-
l-glutamine (all purchased from Biochrom, Berlin, Germany).
Generation of CAR vectors
Cloning and structural composition of conventional CARs as well as the universal chimeric antigen receptor (UniCAR) 28/ζ construct have been described previously in detail [
23,
24,
26]. Briefly, they contain an extracellular binding moiety derived from the αPSCA MB1 mAb [
25], the αCD19 HD37 mAb [
27] or the αLa 5B9 mAb [
28] followed by hinge and transmembrane domain of human CD28 and cytoplasmic activation domains of CD28 and CD3ζ. As a targetable moiety, the peptide epitope E7B6 (E-tag) is incorporated in the extracellular spacer region. The Stop variant of the UniCAR lacks intracellular sequences downstream of the CD28 transmembrane domain.
To generate a UniCAR lacking the E-tag, the DNA fragment coding for the extracellular part of the CAR 28/ζ was cut out of the respective vector with SfiI and HpaI. A DNA sequence encoding the scFv 5B9 and hinge region without the E7B6 tag was synthesized by Eurofins Genomics (Ebersberg, Germany) and cloned via SfiI and HpaI into the expression vector p6NST60-CAR 28/ζ, generating the ΔCAR 28/ζ construct.
The novel αE-tag CAR 28/ζ was obtained as follows: first, a fragment encoding the antigen-binding domain was generated by cutting the sequence of the humanized scFv La (7B6) VL–VH out of a pSecTag2B vector with NheI and ApaI. The orf of a fragment containing the hinge, transmembrane and signaling domain of human CD28 as well as human CD3ζ signaling chain was cut out of a cloning vector available in our lab using the restriction enzymes MssI and ApaI. Finally, the expression vector p6NST60-MCS was digested with XbaI/KspAI and ligated with the aforementioned two fragments resulting in the vector p6NST60-hu αE-tag CAR 28/ζ.
The Orf of all CAR constructs was C-terminally fused to the DNA sequence coding for enhanced green fluorescent protein (EGFP). To allow for a co-translation of CAR and EGFP from one mRNA, both reading frames are separated by a 2A ‘cleavage’ site derived from Thosea asigna virus that induces a ribosomal ‘skip’ from one codon to the next without the formation of a peptide bond [
29].
Production and purification of αPSCA targeting modules (TMs)
Design, cloning, and purification of αPSCA-E5B9 TM from 3T3 cell culture supernatants was previously described in detail [
25,
30]. For redirection of αE-tag CAR T cells, murine or humanized αPSCA scFvs (clone MB1) were C-terminally modified to contain the E7B6 epitope (E-tag). The respective orfs were cloned into the p6NST50 vector which was used to generate a stable CHO cell line for permanent TM production. Purification from cell culture supernatants and analysis of protein concentration were conducted as reported elsewhere [
25,
30,
31].
Isolation, transduction, and expansion of human primary T cells
Buffy coats were supplied by German Red Cross (Dresden, Germany). Human CD3
+, CD4
+ or CD8
+ T cells were isolated and cultured as described previously [
25,
28,
30,
31].
Production of lentiviral particles using HEK293T cells and T-cell transduction was conducted as reported elsewhere [
23,
26,
32]. In brief, T cells were stimulated using αCD3/CD28 DynaBeads
® (Invitrogen, ThermoFisher Scientific) at a 1:4 bead to cell ratio. For T-cell transduction concentrated virus supernatant was added five times within the first 3 days of expansion. After successful genetic modification verified by EGFP expression, cells were sorted on a FACSAria III (BD Biosciences Pharmingen, Heidelberg, Germany). One day prior to experiments, T cells were rested in complete RPMI 1640 lacking any recombinant cytokines.
Flow cytometric analysis
Flow cytometry was carried out on a MACSQuant
® Analyzer (Miltenyi Biotec, Bergisch Gladbach, Germany) and acquired data were analyzed using MACSQuantify
® Software (Miltenyi Biotec) or FlowJo 10.1 Software (TreeStar Inc., Ashland, OR USA). Fluorescently labeled mAbs directed against human CD3 (clone BW264/56), CD4 (clone REA623), CD8 (clone REA734), CD69 (clone FN50), granzyme A (GzmA, clone REA162) and granzyme B (GzmB, clone REA226) as well as respective REA isotype controls (clone REA293) were purchased from Miltenyi Biotec. mAbs against human CD107a (clone H4A3) and IgG1 (isotype control, clone MOPC-21) were obtained from BD Biosciences Pharmingen. For the detection of E-tagged CAR constructs on genetically modified T cells, the αE-tag mAb (clone 7B6) and secondary goat anti-mouse IgG F(ab′)2-PE Ab (Beckmann Coulter, Krefeld, Germany) were used [
23].
After staining of extracellular markers and live vs. dead cell discrimination applying a Zombie Red™ Fixable Viability Kit (BioLegend, London, UK), T cells were processed using an Inside Stain Kit (Miltenyi Biotec) according to the manufacturer’s protocol and subsequently labeled for intracellular markers.
Cytotoxicity assays
The elimination of CAR T cells by αE-tag CAR T cells was assessed using a previously established flow cytometry-based assay [
33,
34]. To distinguish effector from target cells, the latter were labeled with 10 µM cell proliferation dye eFluor™450 (eBioscience, ThermoFisher Scientific) according to the manufacturer’s instructions. The next day, effector cells (αE-tag CAR T cells) were incubated with eFluor™450-stained CAR 28/ζ, ΔCAR 28/ζ or CAR Stop T cells at indicated effector to target cell (
E:
T) ratios. After indicated time points, cocultures were carefully resuspended and an aliquot of 25 µl diluted 1:4 with 1 µg/ml propidium iodide (Sigma-Aldrich, Munich, Germany) was measured at a MACSQuant
® Analyzer. After exclusion of doublets and dead cells, cells/ml were assessed for both effector and target T cells separately. Finally, cells of one triplet were pooled, stained for intracellular expression of GzmA and GzmB, and analyzed by flow cytometry.
The antitumor activity of αE-tag CAR T cells was determined in standard chromium-51 release assays as previously reported [
25].
T-cell activation and degranulation assay
A total of 5x105 eFluor™450-labeled CAR 28/ζ, ΔCAR 28/ζ or CAR Stop T cells was cultured together with equal numbers of αE-tag CAR T cells. Additionally, 5 µl of αCD107a mAb or IgG1 isotype control were pipetted to each well. After 1 h of incubation at 37 °C, 1 µl of 2 mM monensin (Sigma-Aldrich) was added. After 20 h, cells were harvested, stained with αCD69 and αCD3 mAb and measured using a MACSQuant® Analyzer. Living cells were distinguished from dead cells by being propidium iodide negative.
Cytokine release assay
For analysis of TNF, IFN-γ, GM-CSF and IL-2 concentrations in cell-free culture supernatants (stored at − 80 °C) enzyme-linked immunosorbent assay (ELISA) was conducted using OptEIA™ Human ELISA Kits (BD Biosciences Pharmingen) according to the manufacturer’s protocol.
Optical imaging of tumor xenograft mice
Five-week-old male NMRI-Foxn1
nu/Foxn1
nu mice were s.c. injected into their right flank with 5 × 10
5 firefly luciferase (Luc)-expressing PC3-PSCA cells (
n = 5 per group). Control animals received tumor cells only or tumor cells mixed with either 5 × 10
5 CAR 28/ζ T cells or 15 × 10
5 αE-tag CAR T cells. For induction of anti-tumor activity, 10 µg of E5B9- or E7B6-tagged αPSCA TM was administered in addition to tumor cells and CAR-engrafted T cells. To impair the anti-tumor response, one group of mice was transplanted with PC3-PSCA-Luc cells, 10 µg αPSCA-E5B9 TM, 5 × 10
5 E-tagged CAR target T cells, and 15 × 10
5 αE-tag CAR effector T cells. Animals were anesthetized 6 h, 1 day and 2 days post-cell mixture injection and bioluminescence imaging was conducted as described elsewhere [
26,
32]. Optical images were created with Bruker Multispectral software (Bruker, Germany).
Statistical analysis
Statistical significance was evaluated with GraphPad Prism 7 software (GraphPad Software Inc., La Jolla, CA, USA). Statistical tests were used as indicated in figure legends. p values of less than 0.05 were considered significant.
Discussion
Although gene-modified CAR T cells have demonstrated unparalleled antitumor responses in hematological malignancies [
2‐
6], adoptive therapy is often associated with severe, partly life-threatening side effects [
4‐
11]. These side effects can be divided into short-time and long-time effects. For example, soon after the start of a CAR T-cell therapy, the adoptively transferred T cells can rapidly expand potentially leading to a severe cytokine storm. In addition, the simultaneous disintegration of a huge number of tumor cells may cause a tumor lysis syndrome. After removal of the tumor cells still side effects may occur due to cross-reactivity of the anti-tumor domain of the CAR with healthy tissues as even seen with the CD19 CAR. For these reasons, safeguards to control toxicity are required. There are currently two major ideas followed to control such severe side effects: (i) elimination of CAR T cells and (ii) engineering of tunable CAR T cells based on molecular switches and/or combinatorial targeting according to the rules of Boolean algebra [
41,
42]. The activity of switchable CARs can be controlled, e.g., by soluble tumor-specific Ab-based components such as the TMs in the UniCAR system [
24,
26,
27,
32,
35‐
38] or even chemically tuned via small molecules (ON-switch CARs) [
43]. Bearing the high costs of CAR T-cell products in mind, approaches aiming to fine-tune CAR T cells should be favored in the first place. However, it would be advantageous to have an additional safeguard available if switchable CARs do not behave as predicted. Although unlikely for terminally differentiated cells and not observed until now, one should always keep in mind that genetic manipulation of cells harbors the risk to generate leukemic CAR T cells either by accidentally transducing single leukemic cells during the manufacturing process or by insertional mutagenesis. In case such leukemic CAR T cells develop, elimination strategies for both conventional and switchable CAR T cells are highly required.
Here, we report an efficient chance to selectively deplete genetically engineered T cells by autologous αCAR-redirected T cells. For this purpose, we have integrated a specific peptide epitope (E7B6) into the CAR architecture that can be used as an inherent elimination tag. Based on a mAb recognizing this tag, an αE-tag CAR construct was designed and successfully generated. Using flow cytometry-based cytotoxicity assays we demonstrate that T cells equipped with this novel CAR selectively bind and eliminate CAR T cells with an incorporated E-tag whilst CAR T cells lacking this tag are not attacked.
Even though the CD8
+ subpopulation has long been considered as the most potent T-cell subtype in terms of cytotoxicity, our killing data reveal that CD4
+ effector T cells eliminate target cells equally efficient via granzyme-mediated apoptosis. One likely explanation is that our read-out is performed after 24 h and 48 h, whilst other studies perform short-term cytotoxicity assays over 4 h [e.g.,
44]. In that regard, we have previously published that the onset of killing via CD4
+ T cells is delayed, with no obvious effects seen 5 h or 6 h after stimulation. Yet, upon prolonging incubation to 20 h, a substantial cytotoxic activity comparable to that of CD8
+ T cells is detectable [
34,
45].
Our data further indicate that target cells expressing high CAR levels are more prone to depletion by αE-tag redirected effectors than cells with low CAR expression. This is in accordance with previously published studies reporting a positive correlation between the number of EGFR cell-surface molecules and cetuximab-triggered ADCC [
18,
46]. However, in the aforementioned study and other currently investigated safety strategies, the suicide/elimination marker is separated from the CAR construct [
13‐
18,
20]. Given that T cells with low safeguard molecule expression display a reduced susceptibility to apoptosis, CAR-engineered T cells lacking the suicide/elimination marker might emerge due to selective pressure. These cells ultimately escape control and perpetuate toxicity. By contrast, integrating the E-tag into the extracellular spacer region of the CAR should prevent CAR T-cell escape. An additional benefit of this approach is the small size of the peptide epitope of only 18 aa. Opposed to that, other proposed depletion markers are relatively large proteins that add a substantial payload to the expression vector potentially interfering with transcriptional efficiency. Most recently, a study reported the development of a so-called CubiCAR, combining three functions (detection, purification and depletion) in one CAR molecule by incorporating CD20 mimotopes and a CD34 epitope into the extracellular region of a CAR [
47]. Although not investigated in the present manuscript, also the E-tag has utility beyond function as elimination marker. As previously shown, it enables detection, selective ex vivo expansion and purification of E-tagged CAR T cells prior to adoptive transfer [
23]. In that regard, we reason that in vivo tracking of gene-modified cells using for example radio-labeled mAbs against the E-tag is feasible as well.
Despite the fact that all currently available elimination markers, in contrast to suicide genes, possess multifunctional characteristics [
17‐
20,
47], several drawbacks are associated with the use of mAbs. First, the efficacy of CAR T-cell depletion via ADCC might be impaired in heavily preconditioned cancer patients due to their compromised immune system. Second, cetuximab and rituximab inevitably provoke on-target adverse effects upon recognition of healthy epidermal tissue and endogenous B cells, respectively. In this regard, using an epitope derived from the human nuclear La/SS-B protein provides a safe approach as the targeted antigen is not available on the surface of intact cells under physiological conditions [
28,
48]. Third, to ensure therapeutic efficiency, a sufficiently high local Ab concentration is required which might not be achieved in all tissues due to limited penetration and retention of infused mAbs. In light of this and given that T cells elicit highly potent cytotoxic effector mechanisms, αCAR-redirected T cells represent a promising approach to bypass the above-discussed shortcomings of drug-based depletion strategies. In contrast to mAbs, T cells engraft and proliferate, hence, can be considered as self-amplifying “living drugs”. Furthermore, they show better tissue penetration and might, therefore, be superior to mAbs in eliminating CAR T cells. Considering the high costs of CAR T-cell products, clinical application of αCAR T cells as a safety switch may be particularly relevant (i) if other mechanisms/strategies fail to control serious treatment-related toxicities or (ii) if leukemic CAR T cells occur.
Nevertheless, once tumor-specific CAR T cells are completely wiped out, therapeutic antitumor activity cannot be retrieved. However, our proposed safety approach offers the unique feature of repurposing αCAR-engineered T cells as tumor-fighting UniCAR T cells by administration of an E-tag-comprising TM. Thereby, tumor treatment can be re-initiated on demand in case of relapse, which represents a substantial advantage over all currently available safeguard strategies.
In summary, we provide first experimental evidence for using αCAR-redirected T cells for selective elimination of gene-modified T cells in case of life-threatening CAR therapy-related side effects. For specific recognition of CAR-expressing T cells, we integrated a small epitope tag (E-tag) into the CAR architecture without impairing functionality of the original construct. This E-tag then serves as a targetable moiety for specific and stringent CAR T-cell depletion via autologous αCAR-engrafted T cells. As the E-tag can be incorporated into all CARs irrespective of the targeted tumor antigen, it represents a promising universal tool to enhance safety of all kinds of cell-based immunotherapies.
Compliance with ethical standards
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.